How to standardise a model in Onyx

Home Forums Onyx Helpdesk How to standardise a model in Onyx

This topic contains 1 reply, has 2 voices, and was last updated by  Timo von Oertzen 3 months ago.

Viewing 2 posts - 1 through 2 (of 2 total)
  • Author
    Posts
  • #413

    Christopher

    Hello,
    I am a new SEM and Onyx user with a reasonably good understanding of the basics by now.
    I have created a model in Onyx, managed to specify relationships between latent and observed variables, freed and fixed parameters and have checked various indicators of model fit.
    However, I would like to know how I can fully standardise my model.
    I have seen the option to “z-transform” individual variables.
    I am unsure whether I should simply z-transform all my variables so that all coefficients in the model will be standardised (for example, covariances will turn into correlations)? Or is there a different way to do this?
    I have also seen the introduction of constants to standardise latent variables, but to be honest, at this point I am not sure of the role of constants in SEM.
    I would be very grateful for a reply regarding model standardisation. Any comments regarding the use of constants in SEM would also be greatly appreciated.

    #764

    Timo von Oertzen
    Participant

    Hi Christopher,

    sorry for the late reply, your post somehow slipped my attention! I hope I can still be of help.

    Yes, if you z-transform all observed variables, than all arrows between observed variables will give you standardized results; note, however, that this may not be the case for latent variables (which may still have a total variance unequal to one), or (very rare case) your model does not allow enough degrees of freedom to fit all variances to one. An alternative is to right-click on a path and select under “customize path” the option “show standardized estimates”, then in addition to the unstandardized parameter the label on the arrow now also shows (in paranthesis) the standardized estimate, i.e., betas for regressions and correlations for covariance edges.

    The constant is a way to introduce means; by default, all means of the observed variables are either freely estimated (which is called “implicit means” in Onyx) or fixed to zero (which is called “explicit means”). In the explicit means mode, however, you can use the constant to actively estimate means of certain latent or observed variables by connecting the constant triangle to the variable in question by a regression path an freely estimate this path. You could think of the constant as being a variable which always is exactly one, without any variance.

    I hope this helps, cheers,

    Timo

Viewing 2 posts - 1 through 2 (of 2 total)

You must be logged in to reply to this topic.

Posted in